Add agent restart lifecycle command

Extends the existing CheckIn RPC with a one-shot AgentCommand
(restart only -- stop/uninstall need real per-platform OS
service-manager integration and stay deliberately out of scope),
delivered at-most-once: cleared the instant it's handed to the agent
in a response, since a restarting agent's process is gone before it
could ever confirm receipt. On restart, the agent flushes whatever's
buffered, aborts its source task, and exits cleanly, relying entirely
on the host's own service manager to bring it back up.

Issuing a command is gated at RoleAdmin (stricter than config
editing's RoleEditor) and logged into the same audit_log table Phase
7's AI interactions use, via a new agent_command event type.

A real bug was found and fixed during live verification: the first
implementation tried to atomically read-and-clear pending_command in
a single INSERT...ON CONFLICT statement using a sibling CTE
referenced only from RETURNING, on the assumption that Postgres
evaluates every part of a WITH query against one pre-statement
snapshot. That's wrong specifically for FOR UPDATE, which always
reads the latest row version including one written earlier in the
same statement -- confirmed empirically (a restart command was
always coming back empty even when genuinely pending, so the agent
never received it). Fixed by splitting into two real, ordered
statements inside one explicit transaction.

See /docs/agent-management-design.md's "Lifecycle commands" section.

proto/sentry/agent/v1/agent_control.proto

@@ -71,9 +71,39 @@ message DesiredOverride {
   string version = 6;
 }
 
+// AgentCommand is a one-shot action, not a persistent desired state like
+// DesiredOverride -- delivered at-most-once (see CheckInResponse's
+// comment). Scoped deliberately narrow: only RESTART exists today.
+// STOP and UNINSTALL are real, disclosed future work, not oversights --
+// both need genuine OS service-manager integration (systemd's
+// Restart=/RestartPreventExitStatus= semantics vs. Windows SCM recovery
+// options are different enough per platform that hand-waving them would
+// be dishonest), which RESTART doesn't: a graceful shutdown followed by
+// a clean process exit, relying on whatever restart policy the host's
+// service manager already has configured -- the same contract systemd/
+// SCM already expect from any well-behaved service.
+enum AgentCommand {
+  AGENT_COMMAND_UNSPECIFIED = 0;
+  AGENT_COMMAND_RESTART = 1;
+}
+
 message CheckInResponse {
   // False when no override has ever been set for this agent -- it
   // should be running whatever agent.toml already has, untouched.
   bool has_override = 1;
   DesiredOverride override = 2;
+  // AGENT_COMMAND_UNSPECIFIED when there's nothing to do. Unlike
+  // DesiredOverride, there is no "applied_command_version" echoed back
+  // in CheckInRequest: the server clears a pending command the moment
+  // it hands it out in a response (see
+  // ingest/internal/agentregistry.Registry.CheckIn), not once the agent
+  // confirms execution -- a restarting agent's process is gone before
+  // it could ever send that confirmation. This is an honest at-most-
+  // once delivery, not at-least-once: a command lost to a network
+  // failure between this response and the agent acting on it is simply
+  // lost, same as any fire-and-forget signal. Re-issuing (PUT
+  // /agents/{host}/command again) is the operator's recourse, same as
+  // it would be for a `systemctl restart` that silently failed to reach
+  // its target.
+  AgentCommand pending_command = 3;
 }